The invention relates to techniques for removing and analyzing microscopic particles from a sample surface, particularly from semiconductor samples.
In the semiconductor industry, unexpected particles due to contamination will cause yield loss during the manufacturing process. Since a major focus of this industry is aggressive reduction in feature size for pattern line widths, the minimum size of particles that can cause performance loss also decreases rapidly. A reasonable estimate for a xe2x80x9ckiller defectxe2x80x9d size is that greater than one-third the size of the smallest feature on the semiconductor wafer.
Although semiconductor manufacturing is performed in clean rooms with stringent particle standards, unexpected contamination will occur due to sources such as moving parts, human presence, gas condensation, and chamber wear. Control and removal of these particles is a continuous process. In many cases, removal of the source of particles requires an understanding of their origin. Many of these particles or defects are too small for detection in a general purpose optical inspection microscope, so higher resolution methods are required, using charged-particle microscopes, such as scanning electron microscopes (SEM), transmission electron microscopes (TEM), scanning Auger microprobes (SAM), or focused ion beam (FIB) instruments, are required.
Even an image of the particle is usually insufficient to trace the origin of the particle, and more information is required. Elemental composition is valuable in identifying the defect. This can be done in various ways using the charged-particle systems mentioned above. Unfortunately, most of the analytical methods are limited by background signals from the environment of the particle.
Throughput is also a critical parameter in semiconductor manufacturing. Existing strategies for compositional analysis of particles on a semiconductor wafer, for example, usually require removal of the wafer from the fabrication area for off-line analysis using methods such as those described below. Removal from the line severely reduces the throughput of the manufacturing process.
Particle identification on sample surfaces using electron-beam based identification is complicated by the size of the particle relative to the electron penetration depth, and by the nature of surrounding materials in the sample. As an electron beam interacts with bulk solid materials, it expands to fill a teardrop-shaped volume as it loses energy. As the primary beam interacts with atoms in this volume, it generates low energy Auger electrons and X-rays that are characteristic of the elements involved.
The particular X-ray line generated will depend on the atomic number of the element, the energy of the electron during the interaction, and other factors. When trying to identify an unknown particle using conventional Energy-dispersive X-ray Spectrophotometry (EDS), the energy of the electron beam must be large enough to generate inner-shell X-rays from all possible relevant elements, which, for semiconductor applications, may include elements of high atomic number such as tungsten. Unfortunately, this energy results in a penetration depth that may be much larger than the particle of interest, resulting in X-ray generation from the sample surface. These X-rays interfere with any signal from the particle, making unique identification of the particle material difficult. Conventional strategies for solving this problem involve either resolving the X-ray lines of different elements, or reducing the energy of the exciting electron beam.
For example, it is possible to detect and analyze electron-beam generated X-rays from a particle by measuring the intensity and diffraction angle of the X-rays diffracted by a reference crystal, or Wavelength Dispersive X-ray Spectrometry (WDS). One chooses the crystal atomic spacing to deflect (with very high resolution) X-rays of a given energy, thus allowing separation between X-ray lines of different elements. This method has higher energy resolution than EDS but much slower throughput. In addition, if the particle could be, as it often is, of the same composition as the sample surface, this method will not uniquely determine the particle composition.
Other solutions involve reducing the energy of the primary electron beam to guarantee the activated volume is less than the volume of the particle of interest. This reduction in primary electron-beam energy results in characteristic X-rays of much lower energy (M or L shell X-rays, rather than K shell). Conventional cooled semiconductor-based detectors use the generation and collection of electron-hole pairs as a measure of the energy of the ionizing radiation (a few eV for each electron-hole pair, depending on the detecting material). A reduction in the X-ray energy therefore leads to a reduced number of electron-hole pairs and reduced sensitivity to the particle material. In addition the resolution of these detectors is governed by the statistics of the electron-hole generation process, and reducing the energy of the detected X-ray often leads to ambiguous identification of the element of interest. X-ray micro-calorimeter methods have been used to detect these weak X-ray signals, using heat transferred to the detector rather than the generation of electron-hole pairs. This process does allow measurement of small X-ray energies, but micro-calorimeter instruments are expensive, have complicated cooling requirements, and are slow compared to other methods. Also, the electron beam must be kept smaller than the smallest dimension of the particle of interest, rendering the method impractical for small, unsymmetrical particles.
Scanning Auger microprobe analysis also uses an electron beam to irradiate a particle of interest, but rather than detecting any X-rays generated it focuses on the detection of Auger electrons ejected from the atoms of the material. These Auger electrons come from outer shells and have relatively low energies. The Auger electron energies from a material produce a pattern that is characteristic of each element in the material, and the shape and exact energy of the Auger transitions provide information on the chemical bonding of the elements in the material (such as, phase or compound information). The escape depth of these electrons is quite small (a few nm), so Auger analysis focuses mainly on the surface of a sample. This is an advantage for the analysis of small diameter particles ( less than 10 nm). For the analysis of larger particles, one can generate depth profiles by using an ion beam to sputter through the particle and take periodic measurements, but this is inherently destructive of the surrounding sample due to ion milling in the SAM, and requires background analyses on the sample near the location of the particle. Auger analysis is typically more sensitive to light elements than standard EDS analysis, making it more suitable to identify organic materials. However, to improve counting statistics, high electron beam currents are typically employed. This exaggerates the issues of thermo-mechanical drift and drift due to electrical charging of the sample. This means that operating the SAM in the xe2x80x9cspot mode,xe2x80x9d with the electron beam positioned on the particle, involves a risk that over time the electron beam spot will drift onto the sample that surrounds the particle. And the use of a raster pattern for the electron beam will be more tolerant of drift for keeping the beam on the particle, but will involve significant contamination of the results with signal from the surrounding material. In either case, background contamination of the Auger results is a serious issue, and Auger analyses of the surrounding material are required to uniquely identify the signal from the particle. The acquisition of background analyses reduces throughput and inherently damages the sample.
TEM can often be used for analysis of particles in or on surfaces. There are a variety of methods for isolating the particle for analysis, including replication, lift-out or cross sectioning the area of interest. These methods all destroy the sample surface and must be done off-line, thereby increasing cost and cycle time.
Moving the particle from the first sample surface to a more controlled environment for testing can dramatically improve the chance of success and throughput for elemental identification with either EDS or Auger analysis. A critical part of this process is the strategy for moving the particle. This disclosure describes a novel method for removing a particle of interest from a sample surface, transporting that particle to a second sample surface with a controlled X-ray or Auger background, and performing electron beam-induced X-ray analysis or Auger electron analysis there, using any of the methods discussed above. This eliminates the requirement that the analyzing technique have high spatial resolution, although a technique with high spatial resolution, such as EDS analysis in the SEM and SAM analysis, is generally preferred. For example, techniques without high spatial resolution that could be successfully applied to the situation of a particle on a reduced or non-interfering background include X-ray Photoelectron Spectroscopy (XPS) and X-ray Fluorescence analysis (XRF), which may offer an advantage in unique and specific situations.
The proposed method for particle manipulation and EDS X-ray analysis can be done in-line on existing wafer-manufacturing tools. An in-line procedure using existing manufacturing and inspection tools represents a significant reduction in cycle time for contamination removal. SEM is a routine method for wafer inspection, and analytical methods using the electron beam in an SEM system provide a substantial throughput advantage over the off-line strategies.
Although this disclosure primarily illustrates the use of the novel technique to manipulate and examine particles that are contaminants in the context of semiconductor manufacturing, the reader should note that the term xe2x80x9cparticlexe2x80x9d may be taken to include objects that may not be contaminants in other environments, such as chemical deposits, biological material, or micro-mechanical machines. In the latter cases, the novel methods of manipulation described in this application may be applied to manipulate these objects generally, for purposes other than electron-beam X-ray analysis or Auger electron analysis.